katz graduate school of business, university of pittsburgh
TRANSCRIPT
Discounting and underpricing in seasoned equity offers
Oya Altınkılıça, Robert S. Hansenb,*
aKatz Graduate School of Business, University of Pittsburgh, Pittsburgh, PA 15260 bA.B. Freeman School of Business, Tulane University, New Orleans, LA 70115
Abstract Discounting and underpricing spread across most seasoned equity offers in the 1990s and were four to five times higher than in earlier years—particularly for riskier and more difficult to market offers, which were more prevalent. Analyses suggest that expected discounting is a cost of uncertainty about firm value, marketing new shares, and acquiring information that raises the offer price. Stockholders appear to recognize this as they incorporate predictable discounting in stock prices when equity offers are first announced. The surprise component of discounting, which reflects the lead bank’s final adjustment to the offer price after the close of trading the night before the offer, releases information that often causes economically large swings in firm value on the offer day. The evidence points to disparities between the issuer’s closing price and the price suggested in the lead bank’s final order book as a primary source of information. The discount surprise appears to be an effective mechanism used by lead banks to update capital suppliers with that eleventh hour information before they commit their funds. JEL classification: G32; G34 Keywords: Underwriters; Seasoned public offerings; Investment banking; Underpricing July 2002 First submitted July 2001 We thank, for their helpful comments, Jim Angel, Alon Brav, Amy Edwards, Charlie Hadlock, Jeff Harris, John Graham, Campbell Harvey, Naveen Khanna, Pete Kyle, Tim McCormick, Urs Peyer, Mark Shroder, Jonathan Sokobin, Paul Spindt, and an anonymous referee. We also thank participants in presentations of earlier versions of the manuscript to the Economic Analysis Group, Securities and Exchange Commission, Washington, D.C.; to the Economic Research Department, National Association of Securities Dealers, Washington, D.C.; and in finance seminars at Duke University, Michigan State University, Tulane University, and the University of North Carolina at Chapel Hill. *Corresponding author contact information: Tulane University, New Orleans, LA 70115 Tel.: 504-865-5624 E-mail: [email protected]
1
1. Introduction
The discounting of seasoned equity offers has become commonplace and is of a larger order of
magnitude in the 1990s than in earlier periods. Discounting is the ratio of the closing market price the day
before the offer to the offer price. In the 1990s it averaged 3.2%, which often exceeds half the
underwriting syndicate’s fee and aggregates to over $2.6 billion. Explanations for discounting of
seasoned equity offers remain limited and untested.
Our investigation divides discounting into two parts. The first part is expected discounting, which is
the predictable component at the close of trading the day before the offer begins. Because discounting and
undepricing share the same expected value (underpricing is the ratio of the offer-day close to the offer
price), to model expected discounting, we rely on the extensive research done on expected underpricing in
unseasoned offers of common stock. The second part of discounting is the eleventh hour surprise that is
revealed when the offer price is made public. In seasoned offers both the eleventh hour discount surprise
and the offer-day return are observable, whereas neither can be seen in unseasoned offers. While the
discount surprise and the eleventh hour unanticipated underpricing are zero on average, they are not equal
when the discount surprise is informative and thus impacts the offer-day return. We investigate this
impact and assess the lead bank’s advantage in gathering private information just before the offer. We
also examine the relation between the offer-day return and the unanticipated underpricing.
Because discounting and underpricing are analogous, we focus on the application of a number of
stories for underpricing in unseasoned offers to explain discounting in seasoned offers. The stories we
examine tend to differ in their assumptions about the source of the lead bank’s information and about how
the bank uses the information. Consequently, they predict different behavior of expected discounting. In
the value uncertainty story, investors should receive more compensation in the form of underpricing as
valuing the firm becomes more difficult (e.g., Rock, 1985). Similarly, we expect lead banks in seasoned
offers to increase discounting as valuing the issuer becomes more difficult. In the placement cost story, as
the offering becomes more difficult to place, higher discounting is needed to pay for attracting capital
suppliers and compensating them for bearing the burden of more illiquidity of longer term investing.
2
Regarding the information acquisition cost story, Benveniste and Spindt (1989) suggest that lead banks in
unseasoned offers improve the offer price by using allocations of more deeply underpriced shares to pay
for information provided by better-informed investors. Therefore, discounting in seasoned offers should
be higher as more positive private news about share price is released. Regarding the rent expropriation
story, Loughran and Ritter (2002) suggest that lead banks in unseasoned offerings exploit gains in issuer
value by allocating more deeply underpriced shares to customers who are likely to repay the bank by way
of future reciprocal deals. Applying this reasoning to seasoned offerings predicts that discounting is
higher when more good news is released about the firm, regardless if the news is public or private.
The initial model for expected discounting contains six variables that are identified from previous
empirical models of expected underpricing in unseasoned offers. Expected discounting in seasoned offers
is bigger for issuers with higher stock return volatility, and for those listed on Nasdaq. It is smaller for
issuers with higher priced stock, and for issues having a more reputable bank leading the underwriting
syndicate. The estimates imply that expected discounting is U-shaped, reflecting a tradeoff as issuers raise
more capital; it falls as proceeds increase keeping the relative amount fixed, and rises for relatively larger
offers, keeping the amount fixed. These findings are robust as each variable is highly statistically
significant. They agree with empirical models of expected underpricing in unseasoned offers, in which the
rationale for these same determinants is that they represent the uncertainty about firm value and
marketing effort.1
To investigate the economic relevance of the pricing stories for seasoned offers, the empirical model
is enlarged to test the response of expected discounting to the release of information during the
registration period (from the filing day through the day before the offer). The findings show that expected
discounting increases with the release of more positive private information during the registration period.
This agrees with the information acquisition cost story. Lowry and Schwert (2001) show that
1 Empirical models of expected underpricing in unseasoned offers are reported in Barry, Muscarella, Peavy, and Vetsuypens (1990), Barry, Muscarella, and Vetsuypens (1991), Hanley (1993), Jegadeesh, Weinstein, and Welch (1993), Dunbar (1995), Beatty and Welch (1996), Booth and Chua (1996), Habib and Ljungqvist (2001), Hansen (2001), Lowry and Schwert (2001), Benveniste, Wilhelm, and Yu (2002), and Busaba, Benveniste, and Guo (2001).
3
underpricing in unseasoned offers rises with the amount of positive private information revealed and is
asymmetrically zero for offers lacking a positive information release. We show that discounting rises with
the release of more negative information, which is compatible with the finding that discounting increases
as volatility rises and thus agrees with the value uncertainty and placement cost stories. We do not find a
significant relation between expected discounting and public information released during the registration
period for the seasoned offers, whether the news is bad or good. This is contrary to the notion that the
banks use discounting to extract rents when the value of the issuer is appreciating. Thus, both seasoned
and unseasoned offers have a number of common factual regularities in the predictable components of
expected underpricing.
The expected discounting model throws light on a long-standing question do investors anticipate
underpricing before equity offers? Rational investors should incorporate the expected dilution cost from
discounting when they evaluate the announcement of the new financing. The price reaction to the
financing announcement indicates that investors anticipate discounting costs, as it is more negative when
predicted discounting is larger. The model also addresses the question we raise, why has discounting
become more common in seasoned offerings? Comparisons of discounting in a sample of 1980s issues
with their 1990s predicted discounting suggest that the average level of discounting has risen in later
years primarily because of a greater propensity of use and increased uncertainty among issuing firms.
The offer pricing stories predict different behavior of discounting’s eleventh hour surprise. While the
eleventh hour discount surprise could simply be a random last-minute price adjustment, it also could
release information held by the lead bank. If the surprise conveys new information, then the offer day will
be characterized by larger than normal stock returns, both good and bad. Consistent with this, the data
indicate unusually large revaluations of issuing firms on the offer day, but not on the surrounding days.
For instance, issuers have a one-in-five chance of an abnormal 1.2% rise in their stock price and a similar
chance of a 3.1% drop. These unusual and large swings in value often exceed the underwriter’s
compensation. Evidently, investors in seasoned offers face significant offer-day uncertainty about firm
value and issuers face significant uncertainty about the amount of proceeds.
4
To further understand the behavior of eleventh hour pricing, the response of the offer-day abnormal
return to the discount surprise (hereafter sensitivity) is examined. The offer pricing stories predict
different sensitivities. In the placement cost story, eleventh hour discounting resolves the uncertainty
about the gap between the prior day’s closing price and the price implied in the final order book, and it
reflects related changes in placement cost.2 This predicts that discounting changes exceed return changes,
hence a negative sensitivity above minus one. In the information acquisition story, last-minute discounts
are used to buy better information. This predicts that return changes exceed discounting changes, so the
sensitivity is below minus one, but only when the news is good. In both of these stories the bank makes
no other use of the information. In the rent expropriation story, while the source of information is not
material, the bank should use the eleventh hour discount to exploit its information advantage to fine tune
its rent extraction, raising the discount if the news is good and scaling back the discount if the news is
bad. This predicts that return changes exceed discounting changes and sensitivity below minus one,
regardless of the news. Cross-section estimates reveal that an unanticipated 1.0% increase in the discount
is associated with an abnormal 0.62% decline in the stock price. This is consistent with the placement cost
story, and it rejects the prediction from the rent expropriation story. Although it contradicts the
information acquisition story, we report further tests that reveal that the sensitivity is more negative, but
still above minus one, when offer-day news is positive. These results concur with the conclusion that
placement cost is the primary use of eleventh hour discounting and that eleventh hour information
gathering may play a secondary role.
The evidence shows that unanticipated underpricing is negatively correlated with the offer-day return.
This agrees with the placement cost story; disagrees with the rent expropriation story, which predicts
positive correlation; and is inconsistent with the information acquisition story. However, our tests show
2 For a discussion of the order book, and investors’ indications of interest and price and quantity bids contained therein, in the case of unseasoned offers, see Cornelli and Goldreich (2001) and Sherman and Titman (2002). In addition to the order book and informed investors, the eleventh-hour information source could be firm managers. A number of underpricing stories in unseasoned offers assume managers use underpricing to signal their knowledge that the firm is undervalued, underpricing more as the undervaluation increases. We focus on the lead banks having the eleventh-hour information, based on later results that seem to rule out that managers use discounting to signal that their firm is undervalued or that the information may be stored up from prior periods.
5
an asymmetric correlation between the return and unanticipated underpricing, which is consistent with a
secondary role for unanticipated underpricing as a payment for better information in the eleventh hour.
Our findings provide an indirect indication of how lead banks price offers. Theoretical and empirical
papers have suggested that lead banks tend to behave credibly and certify the fairness of offer prices to
maintain their reputations with capital suppliers and thus their lead bank role in future offers (Booth and
Smith, 1986; Smith, 1986). Our results appear to agree with the notion that lead banks possess
economically significant information that investors do not have in the eleventh hour—information that
primarily reflects discrepancies between the market price and the price implied by their order book. The
evidence also indicates that the banks convey this information in the final offer price adjustments. It thus
looks as if alert capital suppliers in all likelihood require the banks to make fair pricing adjustments, and
the banks respond by using discounting as an effective information updating mechanism, before those
investors commit their funds. In agreement with this interpretation, in our study no evidence supports the
application of the rent expropriation story to seasoned offers.
Our study updates the evidence of discounting in the 1980s documented by Loderer, Sheehan, and
Kadlec (1991). The finding of a negative relation between the offer-day return and discounting confirms a
similar finding by Lease, Masulis, and Page (1991) for equity offers during 1981–1983.
The remainder of the paper proceeds as follows. Section 2 discusses the sample and the identification
of the offer day. Section 3 discusses underpricing and discounting, investigates the expected discounting,
and provides comparisons with discounting in the 1980s. Section 4 reports the offer-day information
release, examines the return sensitivity to the unanticipated discounting, and assesses unanticipated
underpricing. Section 5 concludes the paper.
2. The sample
The sample of seasoned offers is obtained from the population of seasoned offers reported on the
April 1998 Securities Data Company Worldwide New Issues Data Bases. The Securities Data Company
provides offer data (type; lead bank identity; underwriting contract form; offer date; gross proceeds,
excluding funds from overallotment option exercise; the portions secondary; offer price; shares issued;
6
lead bank identity; size of the global tranche). The issuer’s exchange listing, standard industrial
classification (SIC) code, stock market prices, shares outstanding, daily trading volume, and daily returns
are obtained from the Center for Research in Security Prices (CRSP) files. Where data are available,
closing ask and bid prices are obtained from the NYSE Trade and Quote Database (TAQ). Institutional
ownership of common stock is taken from Disclosure/Spectrum Institutional 13F Common Stock
Holdings and Transactions. Lead bank reputation (hereafter reputation) is measured using the Carter and
Manaster (1990) ranking of banks that underwrite unseasoned equity offerings and as updated in Carter,
Dark, and Singh (1998). Sample banks, identified from the Securities Data Company as lacking a ranking,
are assigned a reputation of zero. All monetary variables are expressed in January 1998 dollars using the
consumer price index. The sample focuses on firm-underwritten, syndicated offers. Although some results
are reported for offerings by utility firms, the primary focus is on offerings by industrial firms (SIC codes
other than 400s, regulated firms; or 600s, financial firms). Also excluded are shelf offerings and offers
that have warrants or are in a unit offer. Results are reported as data are available. Small offers are deleted
from the sample (those under $10 million). This yields a final sample of 1,703 offerings.
Fewer (64) offers were made in 1990, when the U.S. economy was recovering from recession. Over
the sample period the number of offers per year tends to be higher in later years. Almost two out of three
(65%) offers are by Nasdaq-listed firms. The offers are somewhat evenly distributed over the calendar
year, with fewest in the third quarter (21.1%) and the most in the fourth quarter (28.4%). According to the
two-digit SIC codes, 52 industries are represented, with 60% of the offers from firms in seven industries
(Chemicals and Allied Products, 171; Industrial and Commercial Machinery, Computer Equipment, 94;
Electronic, Other Electrical Equipment, Components, Except Computer Equipment, 89; Oil and Gas
Extraction, 84; Business Services, 79; Health Services, 62; Measuring, Analyzing, and Controlling
Instruments; Photographic, Medical and Optical Goods; Watches and Clocks, 57). The mean gross
proceeds is $85.9 million (median of $57.12 million) and as a percentage of outstanding equity is 21.7%
(median of 19.8%), with both measures excluding exercise of overallotment options.
The Securities Data Company offer date is not an accurate indicator of the actual offer day, so the
offer date is identified with a new procedure. The new procedure relies primarily on the offer date
7
reported in the business press. For each issuer we search the Dow Jones News Retrieval Publications
Library (DJNR) for articles reporting the offer, commonly numbering one or two per issuer. The offer
date is that article’s date if its time stamp is before the close of trading. When the time stamp is after the
close, the next day is designated as the offer date. When no article is found on the DJNR, the Securities
Data Company offer date is used. A second identification procedure draws on the Safieddine and Wilhelm
(1996) method that exploits the fact of an enormous trading volume surge on the offer day. In the sample,
for example, offer-day relative volume (volume and shares outstanding are obtained from CRSP) is more
than a dozen times larger than the average measured over the prior 250 trading days or the 250 trading
days commencing one year later. Relative daily trading volume is examined on 11 trading days, centered
on the Securities Data Company offer date. The offer day is chosen as the day among the 11 days on
which relative volume jumps five times over the prior day’s relative volume, provided it is not more than
twice its prior day’s level.
The two procedures identify the same offer day for 98% of the offers, disagreeing in 36 cases. These
discrepancies are resolved by appealing, first, to articles from DJNR and, second, where needed, to the
relative volume criteria. Using this offer date as a benchmark, Securities Data Company classifies over
50% of the offer dates incorrectly, from the standpoint of the day the stock market is open when the offer
begins.3 Most commonly when the Securities and Exchange Commission (SEC) declares the offer
effective after the close of trading, Securities Data Company reports that day as the offer date, while the
offering actually takes place the next day.
3 Because differences between the closing prices the day before and two days before the offer day are not systematic, not correcting for the right offer day does not appear to introduce a significant effect on the main findings about discounting reported in this study. Further comparisons of the prospectus document dates reported by the Securities and Exchange Commission’s EDGAR, for a sample of offers, with offer dates reported by the Securities Data Company gave inconsistent results.
8
3. An examination of discounting
This four-part section examines sample discounting. Part one reviews the definitions of underpricing,
discounting, and the offer-day return, and it compares their sample behavior with findings reported in
earlier studies. Part two reports the estimates of the empirical model of expected discounting. Part three
uses the model with a sample of 1980s offers to examine the growth of discounting in the 1990s. Part four
investigates whether price reactions to equity financing announcements reflect discounting costs predicted
by the empirical model.
3.1. Underpricing, discounting, and the offer-day return
Underpricing, discounting, and offer-day return, denoted respectively by U, D, and R, are measured
as logarithms of the following respective ratios: the offer day’s closing price, p1, to the offer price, po; the
prior day’s closing price, p-1, to the offer price; and the offer-day close to the prior day’s close. From the
identity
×
=
−
−
1
1
o
1
o
1pp
pp
pp
, (1)
taking logarithms shows that underpricing is the sum of discounting and the offer-day return
U = D + R. (2)
Panel A of Table 1 reports mean underpricing, discounting, and offer-day return for issuers listed on
the NYSE or Amex exchange and on the Nasdaq exchange. Underpricing and discounting are higher for
Nasdaq issuers. Reflecting the identity between underpricing and the sum of discounting and the offer-
day return, combined mean underpricing (2.58%) is similar to the sum of the combined mean discounting
(2.47%) and the mean offer-day rate of return (-0.03%).
[Insert Table 1 near here.]
9
The underpricing and discounting percentages in Panel A are economically significant. For the typical
offer, the nominal value for discounting is $1.4 million. Moreover, over 75% of the offers experience
some discounting, averaging 3.2%. The significantly large and widespread discounting rejects the
hypothesis that seasoned offers are priced at the prior day’s closing price. That being said, many offers
are not discounted.
Although underpricing and discounting are larger and common, the mean offer-day return remains
modest. Yet, significant volatility is present in offer-day prices, as is evident in the return standard
deviation of 4.7%, which is 50% above its mean of 3.2% over the prior five trading days and 100% above
its mean of 2.3% over the subsequent five trading days (not reported). While at first it may seem unusual
that the offer-day return is modest, particularly in light of the apparent increase in discounting, it is not
surprising if investors rationally assess possible offer-day closing price states and thus value issuers fairly
just before the offer day.
To place 1990s seasoned offer pricing in perspective, Panel B has mean pricing measures from prior
studies, sorted lowest to highest and by exchange listing. Most offers in these studies occur before 1985.
Based on the Panel B means, underpricing and discounting in the 1990s are four to five times higher than
in earlier years. Not reported in the table is a rising trend in the 1990s discounting, based on its three-
month moving average. However, this is due to a rise in the monthly moving average of the fraction of
offers that are discounted (annual means increase steadily from 70% to 86%), while the trend in
discounting paid in the discounted offers has remained flat (annual means hover around 3.2%). Thus,
while discounting became more frequent, users encountered similar discounting rates throughout the
1990s.4
Fig. 1 depicts underpricing, discounting, and offer-day return frequencies. Fractiles are formed by
rounding to whole points. Thus, for example, the 1% discounting fractile contains all discounting above
0.5% to 1.5%. Underpricing in Panel A is a mixture of the discounting and return distributions, and it 4 In contrast, underpricing in unseasoned offers rises in the 1990s with a dramatic trebling in 1999 and 2000 (Loughran and Ritter, 2001; Ljungqvist and Wilhelm, 2001). For comparison purposes we are able to confirm the flat trend in seasoned offer discounting in these years, as the mean discounting for the 182 offers from the Securities Data Company that meet sample criteria in 1999 and 2000 is 2.93%.
10
shows both light skewness and a large frequency at zero that are traceable to the discounting. The
distinctive feature of discounting in Panel B is that it is essentially truncated at 0% and right-skewed
(skewness is 2.14). Loderer, Sheehan, and Kadlec (1991) report similarly truncated discounting for offers
during 1980–1984. A similar censoring arises in analyses of underpricing in unseasoned offers in which
negative underpricing is typically set to zero and Tobit regressions are used (Hansen, 2001). Thus,
discounting is modeled using the Tobit regression. The offer-day return in Panel C is roughly
symmetrically distributed around its -0.03% mean (median, 0; skewness, -1.21).
[Insert Fig. 1 near here.]
3.2. An empirical model
The initial model of expected discounting is based on the rich research that models underpricing in
unseasoned offers for determinants of discounting. The model is then extended to assess the extent to
which discounting reflects information released during the registration period.
3.2.1. Discounting and offering characteristics
From Eq. (2), expected underpricing equals expected discounting.
E[U] = E[D] (3)
The estimates obtained from the empirical model of expected discounting thus are likewise estimates
of expected underpricing.
Studies of unseasoned offers report mixed results of the effect of the amount of the offering on
underpricing (Barry, Muscarella, Peavy, and Vetsuypens, 1990; Dunbar, 1995; Hanley, 1993). The offer
amount is measured by the natural logarithm of the gross proceeds. Hansen (2001) reports underpricing
increases with the relative amount of the offer. Relative amount is the gross proceeds with regard to the
market value of equity, measured one week before the offer day. An increase in the relative amount, while
keeping the amount fixed, is likely to raise the uncertainty cost of adverse selection and placement
pressure, given that more capital is being sought relative to issuer size. This would suggest greater
11
reliance on discounting for relatively larger amounts. The offer price inverse, a proxy for uncertainty
about value, has a significant effect on underpricing; that is, underpricing is higher for lower priced stock
(Jegadeesh, Weinstein, and Welch, 1993; Beatty and Welch, 1996). Because the offer price is
endogenous, the inverse of the market price as of five days before the offer day is used. Marketing of
lower priced stock is expected to be more difficult, suggesting greater reliance on discounting. A number
of studies report higher underpricing when stock return volatility is higher (Barry, Muscarella, Peavy, and
Vetsuypens, 1990; Barry, Muscarella, and Vetsuypens, 1991; Jegadeesh, Weinstein, and Welch, 1993).
Volatility is measured with the standard deviation of abnormal returns measured over the 200 days ending
one month before the offer. The Table 1 results suggest that discounting is larger for Nasdaq offers. The
model therefore includes an indicator variable equal to one for Nasdaq-listed firms. A number of studies
suggest that issuers hire more reputable banks to certify the offer price (Smith, 1986; Booth and Smith,
1986; Carter and Manaster, 1990). Empirical studies of unseasoned offers that have variable underwriter
spreads, as seasoned offers, report that more reputable banks are associated with lower underpricing (see,
for example, Megginson and Weiss, 1991; Dunbar, 1995; Booth and Chua, 1996; Carter, Dark, and
Singh, 1998). In contrast, studies of “7% plus” contract users, which predominate in unseasoned offers,
show that underpricing increases when more reputable lead banks are used (Beatty and Welch, 1996;
Hansen, 2001).
Fig. 2 shows the bivariate relations between discounting and the selected variables, where the column
heights report mean discounting by variable quintile in all cases except the Nasdaq listing indicator
variable. Discounting is higher for smaller offerings and for relatively larger amounts. It is also higher for
lower price stocks, for more volatile stocks, and for offerings with the lowest reputation lead banks.
Nevertheless, the largest amounts are associated with discounting above 1.25%, as are the proportionally
smallest offers, the highest priced offers, the least volatile offers, the NYSE firms, and the offers with the
highest reputation lead banks. The relations are not driven by extreme values.
[Insert Fig. 2 near here.]
The discounting model also includes the inverse Mills’ ratio, in anticipation of estimation bias from
censored pre-offer pricing and return measures caused by withdrawn offers (Edelen and Kadlec, 2002).
12
The ratio is calculated using the probability of withdrawal estimated by a Probit model (Heckman, 1979;
Maddala, 1983). Panel A of Table 2 shows that compared to completed offers, withdrawals that meet
sample criteria file after lower abnormal returns and are withdrawn after significantly negative pre-offer
abnormal returns (-17.64% with 83% negative, versus -0.34% with 54% negative). Mikkelson and Partch
(1988) report similar return behavior for filings from 1973 through 1984. Panel B has four withdrawal
probability models. Column 1 shows that withdrawal is likelier for smaller firms and smaller offers and is
unaffected by lead bank reputation or exchange listing. Withdrawal is more likely when pre-file returns
are worse (Column 2), and when either pre-offer abnormal returns or market returns are worse (Column
3) confirming Clark, Dunbar, and Kahle (2002). Column 4 includes the return measures when they are
respectively positive, and zero otherwise, and shows that withdrawal is more likely when news is bad.
These results show that pre-offer pricing and returns in completed offer samples are censored.
[Insert Table 2 near here.]
To address further concern with discounting model misspecification associated with omitted
variables, other possible variables were considered, and each discounting model estimation includes
industry-specific dummy variables, one for each two-digit SIC code, and dummy variables for each
calendar year. These variables allow for a different intercept for each industry and year. Throughout, the
coefficients of these variables are not reported.
We model discounting in a multivariate Tobit regression estimation. Column 1 of Table 3 lists the
base case estimates. Larger offers have lower discounting, while relatively larger offers have greater
discounting. Notice that this indicates a U-shape pattern in the discounting, as the amount of the offer
increases, ceteris paribus. This may reflect scale economies at lower amounts that give way to rising
discounting costs for larger offers. The estimates suggest that the U-shape reaches a minimum at a relative
amount near 17% (0.33/1.94). Discounting increases as issuers’ stock prices decline and as their stocks
become more volatile. Capital suppliers also appear to require less discounting in offers led by more
reputable banks. However, they require almost 0.5% more discounting for offers by Nasdaq firms, even
after controlling other issuer and offering characteristics.
[Insert Table 3 near here.]
13
The base model is expanded in Column 2 to include the inverse Mills’ ratio, estimated from the
Column 4 model in Table 2. The ratio’s negative effect suggests that greater withdrawal likelihood
reduces the discount. For unseasoned offers, Benveniste, Wilhelm, and Yu (2002) report a positive effect
of withdrawal probability on underpricing. Busaba, Benveniste, and Guo (2001) report a negative effect
and suggest that issuers who are more likely to withdraw their offers have more power negotiating with
the lead bank. In our setting, in which lousy pre-offer returns are a key withdrawal predictor, the negative
effect may reflect reduced placement effort and below normal purchase of good information when market
conditions are predicting withdrawal is more likely. The stability of the other coefficient estimates in the
discounting model indicates that the estimation bias caused by withdrawals is negligible.
To address the concern that extreme discounting values may be driving the results, the model is
reestimated after winsorizing the discounting; that is, for offers in the lower 5% tail the discounting rate is
set equal to the 5th percentile value of 0%. Discounting rates in the upper 5% tail of the distribution are
set equal to the sample’s 95th percentile value of 8.9%. The winsorized Tobit results, in Column 3 of
Table 3, are essentially the same as those results in Column 1. Column 4 reports qualitatively similar
results that use ordinary least squares, except for the Nasdaq indicator, which is insignificant.
Overall, none of the industry and offer-year dummy variable effects is statistically significant.
Evidently, no particular industry or calendar-year effects are associated with discounting in the 1990s.
Hanley (1993) reports that end-of-quarter institutional holdings of an unseasoned issuer’s common
stock has a positive effect on underpricing, and Hanley and Wilhelm (1995) find institutional holdings do
not have a significant effect on underpricing. We examine the effect of changes in institutional ownership
from before to after the offer day as well as the post-offer level of institutional ownership, constructed
from Spectrum’s files that draw from 13F filings with the Securities and Exchange Commission. Studies
show that underpricing is larger in unseasoned offers by firms in high-technology industries (see, for
example, Mola and Loughran, 2001; Lowry and Schwert, 2001; Bradley and Jordan, 2002) and when
underpricing of other recent unseasoned offers is higher (Bradley and Jordan, 2002; Benveniste, Wilhelm,
and Yu, 2002; Loughran and Ritter, 2002). We look for a high-technology industry effect using hi-tech
industry definitions from these studies (two-digit SICs of 36, 38, 48, or 73) as well as the definition given
14
on the SDC file. We examine the effect of discounting of other recent seasoned issues. We also
investigate the effect of selling a part of the offer in a foreign market and length of time between the filing
date and the offer date. None of these variables is statistically significant, so these findings are not
reported.
We test the prediction of the Gerard and Nanda (1993) model, which assumes that discounting
increases the informativeness of stock price, as measured by increases in offer size relative to trading
volume. No significant relation is found between expected discounting and the ratio of new shares to the
average trading volume over the month before the offer day.
These results show that part of discounting cost is predictable and that important discounting
determinants are the same as those in empirical models of expected underpricing in unseasoned offers.
They are robust to the inclusion of the full range of independent variables and to reasonable methods of
estimation. They are consistent with the use of discounting to address concerns with greater placement
costs.
3.2.2. Discounting and registration period information
During the registration period the lead bank markets the offer to expand demand and gather
information about share value. Theories of unseasoned offer underpricing provide some predictions about
the response of expected discounting to this information. The information acquisition story predicts that
expected underpricing is positively correlated with positive private information revealed during the
registration period. The rent expropriation story predicts that expected underpricing is positively
correlated with both private and public information revealed during the registration period.
To assess whether the registration period information is incorporated in the discounting, three pricing
measures are used. The first pricing measure, called the offer price adaptation, is the ratio of the offer
price to the closing stock price the day after the filing. The price adaptation is similar, but not identical, to
the price revision studied in unseasoned offers, which is the offer price relative to the midpoint of the high
and low price estimates reported in the filing prospectus. Information conveyed by the price revision may
differ from information in the price adaptation to the extent unseasoned offer prospectus prices
15
misestimate market value. The sample mean price adaptation is -1.36% (t-statistic = -3.69), whose
negativity partly reflects the discounting as the mean raw return is 1.60% (t-statistic = 12.48). The second
pricing measure is the return on the market. The return on the CRSP equally weighted index is 1.26% (t-
statistic = 16.25) on average and positive for over 65% of the offers. The third pricing measure is the
issuer’s abnormal return, equal to the difference between raw return and the CRSP equally weighted
return index. The sample mean abnormal return is a statistically significant -0.34%.
Fig. 3 charts discounting versus the respective pricing measures which are sorted in deciles from low
to high. The column heights are within decile mean discounting. Variable means within decile are
reported numerically on the abscissas.
[Insert Fig. 3 near here.]
The discounting-price adaptation relation is a measure of how much bank information is incorporated
in the offer price. A rising discount as the price adaptation increases suggests the bank is not
incorporating all its information in the offer price. Such evidence would be similar to the partial
adjustment phenomenon documented by Hanley (1993), in which underpricing is significantly positively
related to the price revision and thus the bank’s information improves. The discounting-price adaptation
Pearson correlation is a statistically significant -0.17. However, Panel A of Fig. 3 shows a shallow V-
shape pattern that underlies this negative trend. The V-shape pattern agrees with the view that deeper
discounting is exchanged for information when there is more positive information about share price. It
also shows that discounting increases as negative information about share price worsens, perhaps
reflecting more uncertainty and marketing cost for firms that perform more poorly as the offer day
approaches. The relation between discounting and the market return shows how much public information
the lead banks incorporate in the offer price. Panel B shows discounting is not materially affected by
market returns (Pearson correlation coefficient = 0.05), suggesting the offer price reflects all public
information. The discounting-abnormal return relation is a measure of the amount of private information
included in the offer price. Panel C reveals a low V-shape pattern between discounting and the abnormal
returns. This agrees with deeper discounting when there is more positive private information and when
there is more negative private information.
16
To further investigate inclusion of registration period information in the offer price, the price
adaptation is added to the Tobit specification for discounting. Panel A of Fig. 3 suggests the relation is
likely to be asymmetric, so the Column 5 estimates include the price adaptation when it is positive. A
statistically significant asymmetric relation is evident between discounting and the price adaptation in
Column 7. Evidently, discounting is raised further when the price adaptation is more extreme, either
down or up (found by adding the two coefficients).
The discounting-price adaptation relation does not reveal which class of information—public, private,
or both—is being included in the offer price. To allow for asymmetric responses to market information,
the market return and the market return when it is positive are included. The Column 6 estimates show
discounting is unrelated to the public information, whether on the upside or the downside. To allow for
asymmetric response to private information, the abnormal return and the abnormal return when it is
positive are included in the model. The Column 7 estimates show that discounting increases with the flow
of private information, both negative and positive (found by adding the two coefficients). The positive
reaction of expected discounting to positive private information and its insignificant response to public
information agrees with an information acquisition role but disagrees with a wealth expropriation role. Its
response to negative private information agrees with the uncertainty and marketing roles for discounting,
as these offers are likely to be more volatile and thus experience greater placement difficulty, all else the
same.
One concern with the evidence is that there may be cycles in the seasoned offer primary market, like
the cycles in the unseasoned offer market (Lowry and Schwert, 2002), that may induce correlation among
regression errors of offers close in time. To address this concern estimates such as those of Fama and
MacBeth (1973) are examined, by first estimating the Column 7 model in each year of the sample period.
Column 8 of Table 3 reports the average of the year-by-year estimates with p-values based on the
standard deviations of the estimates. The signs of the estimates are consistent with those in Column 7, and
all are statistically significant, except for reputation and the registration period abnormal return.
Another concern is that the results may be driven by the effect of the fraction of the offer that
represents current stockholders’ sales on underpricing. Studies report mixed evidence of the effect of
17
secondary offers on unseasoned offer underpricing (see, for example, Jegadeesh, Weinstein, and Welch,
1993; Ljungqvist and Wilhelm, 2001). To address this concern, separate estimates for the primary and
secondary offers are reported; an offer is primary unless it is 10% or more secondary. The signs of the
estimates, reported respectively in Columns 9 and 10, agree with the Column 7 estimates, and all are
statistically significant except for lead bank reputation and the registration period abnormal return in the
case of the secondary offers.
These additional results therefore indicate that public information is largely incorporated in the
discounting while private information is partially incorporated. They also reveal an asymmetry between
discounting and the release of private information that is consistent with the use of discounting to
compensate investors for new information and for bearing uncertainty and liquidity costs. Comparable
findings are reported for unseasoned offers by Lowry and Schwert (2001). The results do not appear to be
sensitive to censored returns caused by withdrawn offers. Similar results are obtained using other
specifications of the withdrawal choice model and when the inverse Mills’ ratio is excluded altogether.
3.3.Comparison with discounting in the 1980s
With the above empirical model we are in position to examine why discounting in the 1990s is
higher. Using the 1990s sample selection criteria, seasoned offers are obtained from the Securities Data
Company for the five-year period from 1980 through 1984, which is the period of focus of many of the
studies noted in Panel B in Table 1.
Panel A of Table 4 reports means and their differences for selected variables for the 1980s and 1990s
samples. Not reported in the table are sample differences in industry representation. However, industry is
not a significant factor in determining discounting in the 1990s. The mean differences alone do not
provide a particularly obvious explanation for the higher 1990s discounting. The 1990s offers are much
larger and have higher prices (measured in January 1998 dollars), and their banks have higher reputations.
Tobit estimates show that the determinants have disparate impacts. To further assess why discounting is
higher in the 1990s, the predicted discounting for the 1980s offers based on the 1990s discounting model
estimates reported in Column 1 of Table 3 is examined.
18
[Insert Table 4 near here.]
Panel B of Table 4 reports that for all 1980s issuers, their 1990s predicted discounting is almost twice
their actual level, rising from 1.01% to 1.94%, on average, with large increases regardless of exchange.
Table 4 also reports the mean predicted discounting within four levels of actual discounting, ranging from
those who are not discounting types in the 1980s (discounting is under 1%) to the heavy discounters in
the 1980s (discounting is over 3%). Most 1980s offers are not of the discounting type, yet, independent of
exchange listing, their mean predicted 1990s discounting is statistically significantly higher (1.56%) on
average. Furthermore, the offers with higher discounting in the 1980s have 1990s predicted discounting
that is similar or lower, on average. This suggests that discounting is commonplace among most issuers in
the 1990s. It is consistent with a general broadening of capital supplier demand for higher discounting.
The higher 1990s discounting is not due to offer-day misalignment, as this is corrected, or to the increased
proportion of issuers from Nasdaq.
In an attempt to pinpoint shifts in investor demand for discounting, the 1990s Tobit model estimates
are compared with those obtained by fitting the same model to the 1980s sample. The 1980s estimates,
reported in the last column of Table 3, are generally qualitatively similar to the 1990s estimates.
However, some differences in the estimated coefficients suggest that higher discounting in the 1990s
reflects increases in capital suppliers’ need for compensation for uncertainty and illiquidity. In contrast to
the 1990s model, which suggests U-shaped discounting reaches a minimum at a relative amount of 17%,
the 1980s minimum is smaller at a relative amount near 11% (0.16/1.41). Because of the implied increase
scale economies in discounting in the 1990s, discounting is higher for smaller offers. It also suggests
higher discounting for those smaller issuers who seek proportionally more funds in the 1990s, everything
else being the same. While 1990s issuers tend to have higher stock price levels, there is also a tripling of
the stock price inverse coefficient from 5.55 in the 1980s to over 16 in the 1990s. Evidently, capital
suppliers demand larger discounting for issuers with lower priced stock, consistent with a higher premium
for value uncertainty in the 1990s. Volatility is also contributing to higher discounting in the 1990s. This
is partly due to an almost 50% increase in the uncertainty premium for volatility, as indicated by the
higher volatility coefficient (32.49 versus 22.76) and partly due to the fact that issuer volatility is up by
19
50% in the 1990s, on average (see Table 4). Furthermore, greater weight is attached to lead bank
reputation in the 1990s. While reputation does not appear to be important economically or statistically in
the 1980s, suppliers of capital receive more discounting in offers led by less reputable banks in the 1990s.
This result is also evident in Panel F of Fig. 2.5
The evidence suggests capital suppliers in the 1990s tend to require discounting more often and
demand more discounting on riskier and more difficult to market offers. The results point to two other
plausible contributing factors. Greater offer frequency and contracting competition in the 1990s (Gande,
Puri, and Saunders, 1999) complement discounting as an economically desirable underwriting contract
innovation. Consistent with this, underpricing plays a more important role in unseasoned offers in the
1990s (Beatty and Welch, 1996; Hansen, 2001). Gerard and Nanda (1993) suggest another factor may be
the 1988 (SEC Rule 10b–21) prohibition of covering short sales made before the offer with shares bought
at the offer price. To the extent that short sales are an economically efficient mechanism to lower
placement difficulty, their prohibition may have induced a substitution of discounting for short selling.
3.4. Predictable discounting and the offer announcement price reaction
The predictability of discounting suggests that rational investors will anticipate the associated
expected dilution costs when they first learn of the offer. Offer announcements are identified by a search
of the DJNR. Two hundred two offers have other news released on the announcement date (84 are
quarterly earnings reports). Panel A of Table 5 reports, for all offers, that the announcement period
abnormal return is -2.23%, with median -2.24%. Both are highly statistically significant. For the offerings
with no other announcement-day news, the mean is a highly statistically significant -2.32% (t-statistic = -
16.83), with median -2.36%. These abnormal returns are comparable to estimates documented by others.6
5 We also examined underwriting spreads, fitting the ordinary least squares model of Altınkılıç and Hansen (2000), and Altınkılıç (2002), for both the 1980s and 1990s samples. Analysis of the predicted spreads from those models indicates that underwriter spreads are not significantly different across the two samples. 6 Evidence of significantly negative reaction to seasoned offerings is found in Asquith and Mullins (1986), Masulis and Korwar (1986), Mikkelson and Partch (1986), Bhagat and Frost (1986), Barclay and Litzenberger (1988), Hansen and Crutchley (1990), Korajczyk, Lucas, and McDonald (1991), Denis (1991), Choe, Masulis, and Nanda (1993), Bayless and Chaplinsky (1996), and Chaplinsky and Ramchand (2000).
20
[Insert Table 5 near here.]
Panel A of Table 5 reports mean and median announcement abnormal returns by increasing levels of
expected discounting, which is estimated using the Column 1 model from Table 3 and the inverse of the
stock price one week before the announcement. Issuers expecting the least discounting have the highest
abnormal return of –1.11%. Issuers with expected discounting of 1% to 2% have economically and
statistically significantly more negative mean abnormal returns of –2.19%. Mean abnormal returns fall
significantly further to –2.49% for issuers with 2% to 3% expected discounting and statistically
significantly further to –2.83% for issuers expecting the widest discounting. This evidence indicates a
monotonically declining relation between falling abnormal returns and rising expected discounting. It is
not consistent with expected discounting as a positive signal that the firm is undervalued. In signaling
models of underpricing in unseasoned offers better-informed managers use underpricing to signal that
their firm is undervalued (Allen and Faulhaber, 1989; Grinblatt and Hwang, 1989; Welch, 1989).
These results do not hold other factors constant. To address this concern, Panel B reports cross-
section estimates of announcement period abnormal returns. The Column 1 model includes the return on
the CRSP index and the relative amount, which is added to control for offer size effects in the
announcement reaction. Expected discounting has a statistically significant negative impact on the
announcement price reaction. When firm size is added to the model in Column 2, collinearity between
firm size and expected discounting drives the expected discounting coefficient insignificant. In Column 3
a discrete variation of expected discounting is instead used (expected discounting is one when the
predicted discounting is under 1%, two when from 1% to 2%, three when from 2% to 3%, and four for
higher levels). This specification restores the expected discount’s negative effect, while the firm size
coefficient is less significant. Qualitatively similar results are obtained when expected discounting is
measured with a set of dummy variables for the predicted discounting levels and when size is measured
discretely (not reported). While it is unlikely that the multicollinearity problem can be completely
satisfactorily disentangled, the evidence agrees with the conclusion that investors account for expected
discounting costs when they learn of the seasoned offer
21
To identify possible differences between primary and secondary offers, Columns 4 and 5 of Table 5
report their respective estimates. The signs for these estimates are the same as in the Column 3 model and
are generally statistically significant. Investors account for expected discounting costs when they learn of
the seasoned offer, independent of the offer being primary or secondary.
4. Offer-day information release, the return sensitivity, and unanticipated underpricing
This three-part section focuses on the information release of the discount surprise and test predictions
from the pricing stories for the return sensitivity. All variables are as of the offer day, unless noted. Part
one discusses the pricing story predictions. Part two reports tests for the effect of information release on
the offer day. Part three has cross-section estimates of the sensitivity to the discount surprise and the
return-surprise correlations.
4.1. Predicted price effects of the offer-day information release
Deducting Eq. (2) from Eq. (3) yields unanticipated underpricing as the discount surprise plus the
offer-day abnormal return (hereafter return), denoted respectively as µ, δ, and ρ,
µ = δ +ρ(δ) (4)
When the surprise is informative, unanticipated underpricing behavior differs from the surprise,
reflecting the behavior of the return sensitivity.7
While our primary focus is on estimating the return sensitivity, we report the correlation between the
unanticipated underpricing and the return. The pricing stories predict different sensitivity behavior that,
by Eq. (4), translates into predictions of different return correlation with unanticipated underpricing. In
7 The surprise discount dilutes stock price. For example, if it is negative, then the offer price is above its expected level, reducing dilution below what is expected, pushing up the return. To isolate the information release, unanticipated dilution, denoted αδ, is removed from the observed return. The adjustment is most evident in discrete time, noting that the offer day close equals the prior day’s close plus the value of information in the surprise, all on old plus new shares (n + m), less the unanticipated dilution loss on new shares (m). Thus, if i is the information value per share, the end of offer day equity value is p1(n + m) = p-1(n + m) + i(n + m) - (E[po]- po)m. Dividing by p-
1(n + m), the abnormal return is AR = ρ - αδ, and the dilution-free return is ρ = AR + αδ.
22
the placement cost story, if the final order book indicates bad news—that the offer-day price will fall—
then the bank cuts the offer price to clear the book and to cover related unanticipated placement cost
increases. Conversely, if the book suggests good news, then the offer price is raised more than the offer-
day price increase. This predicts negative sensitivity above minus one and negative correlation between
unanticipated underpricing and the return. If the information is purchased from better-informed investors,
when good news is acquired the offer price is raised partially toward the higher offer-day close, so the
return increases more than the discount cut and the underpricing increase. This predicts sensitivity below
minus one and, from Eq. (4), positive unanticipated underpricing-return correlation. In the rent
expropriation story, the bank uses its information advantage, raising the offer price partially toward the
offer-day close if news is good and cutting the offer price less than the drop in the offer-day close if news
is bad, to fine tune its rent expropriation. This predicts sensitivity below minus one and positive
unanticipated underpricing-return correlation, whether the news is good or bad.
4.2. A test of offer-day information release
To test if an unusual amount of information is released on the offer day, the variance ratio is
examined. The variance ratio is an F-statistic that tests if the daily return variance expands significantly
above normal. An issuer’s daily variance ratio is the squared common stock daily abnormal rate of return
relative to its normal benchmark daily abnormal return variance. The mean variance ratio on each of the
11 days, centered on the offer day, is examined. The benchmark variance is estimated with the available
returns over the 200 days ending one month before the offer. Given the large sample size, in the absence
of significant information release, the expected ratio approaches one. Under the information release story,
the offer-day variance ratio should exceed one and the ratios on other nearby days.
Column 1 of Table 6 has the mean variance ratios. The pre-offer ratios hover near one until two days
before the offer, then rise a statistically significant 15% two days before and 26% one day before the
offer. This agrees with the arrival of an unusual increase of information on each day. On the offer day the
ratio rises sharply by 70%. This increase is significantly above both one (t-statistic = 10.64) and the
23
unusually high ratios on the two prior days. The ratios in Columns 2 and 3 show that the information
surge is independent of the issuer’s exchange listing.
[Insert Table 6 near here.]
The variance ratio is significantly below one on each of the five days after the offer, regardless of
exchange listing. This is consistent with a number of offers being stabilized during the syndicates’
placement efforts, which will drive their abnormal returns toward zero, everything else being the same.8
To investigate this stabilization hypothesis, the variance ratio is reestimated on each relative day only for
those issuers having positive returns that day and are thus not materially influenced by stabilization. To
remove the bias from conditioning on a positive return, for each issuer the benchmark variance is
remeasured using only its positive return from the 200 days ending one month before the offer.
Columns 4 to 6 report variance ratio performance for issuers having positive returns. The ratios hover
near one every day around the offer and show a very large statistically significant increase on the offer
day, rising more than 70% on each exchange. They are consistent with unusually large information
release on the offer day. They do not show unusual decline after, consistent with the stabilization
hypothesis.
4.3. The return-discount surprise sensitivity
Although the evidence thus far does indicate unusually high information flow on the offer day, it does
not reveal whether the offer price releases information or the nature of that information. High information
flow could simply reflect the trading activity of informed investors.
We initially estimate the sensitivity using an ordinary least squares model in which the dependent
variable is the return after the adjustment for the dilution from the discount surprise. The primary proxy
for the surprise is the residual from the empirical discounting model in Column 7 in Table 3, which
8 Wilhelm (1999) and [Aggarwal (2000)] report that lead banks in unseasoned offers never use formal, or pure, stabilization. This casts doubt on the use of pure stabilization in seasoned offers. However, as they note, it does not rule out other stabilization-related activities that include short selling and the use of penalty bids.
24
incorporates the latest stock price information. A positive residual is a measure of a deeper than expected
discount (the offer price is below what was expected). A negative residual indicates smaller than expected
discounting. Fig. 4 is a plot of the offer-day rate of return versus the surprise. The return and the surprise
appear to have a strong inverse relation.
[Insert Fig. 4 near here.]
The regression model includes three controls that may affect the return. The first control is
contemporaneous movements in the market, as measured by the return on the CRSP equally weighted
daily index. The second control, the market value of common stock before the offer, registers cross-
section differences between issuer returns that are the result of firm size. The third control is the ratio of
offer-day trading volume relative to benchmark average daily trading volume, which is measured from
the beginning of the 25th week after the offer through the 100th week after the offer. The variable
attempts to control for selling pressures on offer-day stock price that arise from the equity sale that can
reduce the return. Although firms may voluntary withdraw their offers infrequently in the eleventh hour,
we report results that include the inverse Mills’ ratio to correct for possible bias caused by such
withdrawals.
The significantly negative sensitivity estimate in Column 1 of Table 7 indicates that a 1% discount
surprise reduces the return by -0.62%. This estimate shows that positive discounts are bad news, negative
discounts are good news, and the discount surprise is larger than the return response (in absolute values).
It confirms the previous evidence that offer pricing is highly informative and indicates that a major
component of the information is the discount surprise.
The –0.62 sensitivity estimate in Column 1 of Table 7 is significantly above minus one (t-statistic = -
10.84). This concurs with the conclusion that unanticipated underpricing is used primarily to pass through
eleventh hour price changes and meet unanticipated placement costs. Because the sensitivity estimate is
not below minus one, it is contrary to both the information acquisition cost and the rent expropriation
stories for offer pricing. The market return and offer size have positive effects on the return. While the
selling pressure variable has the expected sign, it is not statistically significant using a two-tail test. We
also examined other controls for possible selling pressures, including the relative amount and the number
25
of new shares issued relative to the benchmark average daily trading volume. These variables were also
statistically insignificant.
[Insert Table 7 near here.]
Because the estimated negative sensitivity estimate is above minus one, Eq. (3) suggests a negative
relation should exist between unanticipated underpricing and the return, ceteris paribus. This is confirmed
by their Pearson correlation coefficient of -.37 (p-value = 0.0001). These results agree with the notion that
the offer price is overadjusted in the eleventh hour to the change in the offer-day price in seasoned equity
offers. Furthermore, they are not consistent with unanticipated underpricing providing a positive signal
that the firm is undervalued.
One important concern with the sensitivity estimate is estimation error. Perhaps the first-stage
discounting model estimation generates the negative sensitivity in the second stage. To address this
concern, discounting is investigated as a proxy for the surprise. The estimated sensitivity reported in
Column 2 of Table 7, -0.57, is highly statistically significant (t-statistic = –17.02) and shows that the
sensitivity is not generated by econometric forecast errors. The robustness of the estimate is also evident
in the Column 3 sensitivity estimate that is obtained using the discounting surprise drawn from the Fama
and MacBeth averages reported in Column 8 of Table 3. That estimate, –0.59 (t-statistic = –15.16), is
virtually identical to those reported in Columns 1 and 2.
A second concern is with the nature of the information release. Is the information from another time,
before the eleventh hour? Is it driven by a small subset of the issuers? Is it confined to the offer day?
While the findings suggest that the information is not likely to come from the registration period, perhaps
the lead bank has retained private information about the firm that it learned during the investigation of the
issuing before the registration period. The drop in stock price when firms announce equity offers reflects
concern that managers have hidden information that the firm is overvalued (Myers and Majluf, 1984).
That information disparity may not be fully resolved by the offer announcement, and the lead bank’s
investigation may yield more residual information. Arguably, the bank may wait to release that
26
information in the offer price.9 To test if the information is from the announcement period we assume that
a given discount surprise evokes a greater return reaction, as investors perceive lower adverse selection at
the announcement. Thus, as investors are more confident at the announcement of low adverse selection
costs, the return is more sensitive to a given discount surprise. This suggests that lower adverse selection
likelihood at the announcement is associated with higher offer-day sensitivity.
We employ two proxy variables for adverse selection perceived at the announcement. The first proxy
variable is firm size. Larger firms are monitored more effectively by capital market participants and are
perceived to have lower potential adverse selection. The second proxy variable is the announcement-day
abnormal return. Studies suggest that a worse abnormal return indicates anticipation of worse adverse
selection costs (Asquith and Mullins, 1986; Masulis and Korwar, 1986; Denis, 1991; Bayless and
Chaplinsky, 1996). Column 4 of Table 7 has the sensitivity estimate when the equation is augmented with
the product of the discounting surprise and firm size. The hypothesis that larger firms have greater
sensitivity to the surprise is rejected. When the logarithm of firm size is instead used, which weights
smaller firms differently relative to larger firms, similarly insignificant results are obtained (not reported).
Estimates when the return equation is augmented with a multiplicative term between the discounting
surprise and the announcement-day abnormal return are in Column 5 of Table 7. The results using the
multiplicative term of discounting times the abnormal return are also insignificant.10
9 Although underwriters may release information when it is found, investors and regulators may view such releases, particularly of speculative or evolving information, with skepticism and perhaps as price manipulation. Issuers may also prefer that details of information remain private, knowing the lead bank must price the issue appropriately with regard to that information. Release via the offer price is free of these concerns and is credible as banks stake their reputations on the offer price, accepting legal liability for its fairness. Underwriter trading in the issuer’s shares may release the information, but this is prohibited just before the offer by law (Regulation M after April 1, 1997, and Regulation 10(b)–6, before). Analyses of underwriter risk bearing and selling effort when the issuer delegates the offer pricing decision to the underwriter are examined by Mandelker and Raviv (1977), Baron and Holmström (1980), and Baron (1982). 10 To further test if the information source is tied to announcement date information, we conduct sensitivity tests for regulated utility firms. A number of studies show that utility offer announcements have inconsequential stock price reactions (Smith, 1986). This is because utility firms are not likely to be associated with much information asymmetry (Smith, 1986). The estimates show that the return-discount surprise sensitivity is fully present for offers by utility firms (not reported).
27
To address the concern that extreme values drive the sensitivity, the sensitivity is stratified using a
dummy variable specification of the surprise. The offers are sorted from lowest to highest surprise into
quintiles, then they are formed into four surprise dummy variables, each equaling one only for offers
falling in quintiles I, II, IV, and V. The mean returns in the respective surprise quintiles span a full 5%,
decreasing monotonically, 2.04%, 1.29%, 0.76%, -0.79, and -3.36%, and are statistically significantly
different from zero and from each other.
To address the concern that the sensitivity is not confined to the offer day, the surprise dummy
variable specification is estimated separately on the offer day and on the two days just before and after.
Columns 6 through 10 in Table 7 contain the estimates. On the offer day, Surprise I issuers have the
lowest discounts relative to what was expected. Hence, their surprise should be good news. Consistent
with this, the estimates show they experience an economically large 1.21% (t-statistic = 3.81) increase in
stock price on the offer day. At the other extreme, Surprise V issuers have an economically significant
drop of 3.13% (t-statistic = -10.19) in offer-day stock price. More generally, the dummy variable
estimates show the daily return is monotonically related to the surprise, but only on the offer day. A
reasonable estimate suggests issuers face offer-day dilution adjusted value changes that range in excess of
4.34% of firm value, 40% of the time. On other days around the offer day, the daily return is
unresponsive to the surprise (similarly insignificant unreported results are found on earlier and later days).
This suggests the surprise is reliable news only on the offer day and complements the variance ratio
evidence of a large information release on the offer day. This evidence also shows that the sensitivity
estimate is not biased downward because the surprise is anticipated, or leaks out, as the offer day nears.
There is also a significant weakening of the size and market return coefficient estimates after the offer.
We suggest these results may be the product of stabilization activity.11
11 We investigated several specification issues that could potentially cause the sensitivity. We find all such issues to be unimportant. Using a Tobit procedure to estimate the regression yields a sensitivity of –0.65. Separately estimated sensitivities for NYSE/Amex and for Nasdaq issuers are –0.66 and –0.61, respectively (t-statistics = –7.72 and –10.28, respectively). A similar sensitivity estimate is obtained if we instead use the residuals from the ordinary least squares model in Table 2 or when we use the residual from the Yeoman (2001) formulation for discounting. After controlling for the order flow imbalance using the method of Lease, Masulis, and Page (1991) for all issuers for which data are available from TAQ, the sensitivity estimate is -0.61, which is similar to the sensitivity using CRSP closing prices.
28
Although the sensitivity estimates are contrary to both the information acquisition cost and rent
expropriation stories, the placement cost and information acquisition stories need not be mutually
exclusive across the entire sample. Because the information acquisition story predicts asymmetric
sensitivity behavior, we can test if it has some influence on eleventh hour pricing. To test for an
asymmetric sensitivity the regression is augmented to include two independent surprise variables, the
surprise and the surprise when it is good.
[Insert Table 8 near here.]
The sensitivity estimates in Column 1 of Table 8 indicate that a 1% discounting surprise when bad
news is revealed reduces the return by 0.55% (t-statistic = -11.18), and when good news is revealed the
sensitivity expands to –0.84 (-0.55 – 0.29, t-statistic = -7.59), which is significantly above minus one (t-
statistic = 1.82). These negative sensitivities above minus one suggest that unanticipated underpricing and
the return are negatively correlated whether the news is good or bad, ceteris paribus. This is confirmed by
their Pearson correlation coefficients of -.09 (p-value = 0.01) when good news is revealed and -.49 (p-
value = 0.0001) when bad news is revealed. This evidence is consistent with the placement cost story.
However, the significantly larger correlation (in absolute value) when good news is revealed agrees with a
secondary role for unanticipated underpricing to acquire information. It does not agree with the view that
the unanticipated underpricing is a signal of firm value.
In Columns 2 and 3 of Table 8 the asymmetry tests are reported for the primary and secondary offers,
using their respective surprise estimates from Columns 9 and 10 in Table 3. The sensitivities estimated
within the primary and secondary offer samples are also qualitatively similar to those in Column 1.
For some offers, the lead bank may engage in stabilization activities that prevent worse offer-day
returns. This would suggest the sensitivity estimate is biased upward. To control for this possible effect,
the sensitivity is reestimated for samples that are successively less likely to have stabilization. That is, as
the gap between the market price and the offer price widens. Columns 4 through 6 of Table 8 report
sensitivity estimates as increasingly less underpriced offers are excluded from the sample. Specifically,
offers are first excluded if their unanticipated underpricing is at or below -2.0%, then at or below 0.0%,
29
and then at or below 2.0%. The sensitivity estimate is not materially affected by the removal of these
offers.12
Possible censoring of pre-offer returns caused by withdrawn offers appear to have an inconsequential
impact on the estimates in Tables 7 and 8. Those estimates remain virtually unchanged when using other
withdrawal choice models, as well as when no correction for estimation bias is performed altogether.
5. Conclusion
Discounting of the offer price in firm-underwritten seasoned equity offers is economically large and
common, remaining stable around 3.0% throughout the 1990s. The increased use relative to earlier
periods appears to reflect both greater demand by capital suppliers and higher risk profile of issuing firms.
Our investigation partitions discounting into its expected and surprise components. Estimates from a
model of expected discounting agree that placement costs per dollar raised decline for larger firms given
the amount raised and increase with the amount keeping firm size the same. Discounting is higher for
issuers with lower stock prices and for those with greater stock return volatility. Offers by Nasdaq firms
have larger discounting than NYSE/Amex firms. Discounting is less for banks with better reputation.
These findings are qualitatively comparable to evidence reported in earlier studies of underpricing in
unseasoned offers. Collectively, these results agree with the notion that two important purposes of
expected discounting in seasoned offerings are to compensate investors for uncertainty about firm value
and placement costs. Further evidence backs the view that expected discounting is also a method of
payment for positive information about issuers. The findings do not support the application of the rent
expropriation story to explain discounting of seasoned offers or the idea that discounting represents a
signal by firm managers that their firm is undervalued.
12 We investigate and reject several alternative explanations for the sensitivity findings. It is not driven by new firm-specific information that is released to the public between the close the day before and the offer-day close. A number of tests rule out a widespread tendency for some lead banks to hesitate setting the offer price until after offer-day stock prices have moved closer to the close. For example, discounting does not mirror the return. Also, after dividing the sample of issuers for which opening and closing prices are available from TAQ, into the night group (i.e., those having their offer price reported after the close the night before the offer day) and the remaining day group, both groups have highly significant sensitivity and more so for the night group. Further, there is large and significantly negative return-surprise sensitivity in the overnight return.
30
On the offer day, returns are unusually volatile and significantly negatively related to the discount
surprise. This shows that lead underwriters release economically significant misvaluation information
with the offer price, before capital suppliers buy new shares. The return-surprise sensitivity estimates
concur with the idea that the discount closes the gap between the issuer’s prior closing price and the price
for new shares implied by the final order book and adjusts for related changes in placement difficulty.
The findings suggest that information acquisition may play a secondary role in the eleventh hour pricing.
That the banks price the offers fairly and do not show a tendency to use discounting to extract rents from
issuing firms is supported by the evidence.
References
Aggarwal, R., 2000. Stabilization activities by underwriters after initial public offerings Journal of
Finance 55, 1075–1103.
Allen, F., Faulhaber, G. R., 1989. Signaling by underpricing in the IPO market. Journal of Financial
Economics 23, 303–324.
Altınkılıç, O., 2002. An examination of seasoned equity offer placement effort. Unpublished manuscript,
University of Pittsburgh.
Altınkılıç, O., Hansen, R. S., 2000. Are there economies of scale in underwriting fees?: evidence of rising
external financing costs. Review of Financial Studies 13, 191–218.
Asquith, P., Mullins, D., 1986. Equity issues and offering dilution. Journal of Financial Economics 15,
61–89.
Barclay, M. J., Litzenberger, R., 1988. Announcement effects of new equity issues and the use of intraday
price data. Journal of Financial Economics 21, 71–99.
Baron, D.P., 1982. A model of the demand for investment banking advising and distribution services for
new issues. Journal of Finance 37, 955-976.
Baron, D.P., B. Holmström, 1980. The investment banking contract for new issues under asymmetric
information: Delegation and the incentive problem. Journal of Finance 35, 1,115-1,138.
Barry, C., Muscarella, C. J., Peavy, J. W., III, Vetsuypens, M. R., 1990. The role of venture capital in the
creation of public companies. Journal of Financial Economics 27, 447–471.
Barry, C., Muscarella, C. J., Vetsuypens, M. R., 1991. Underwriter warrants, underwriter compensation,
and the costs of going public. Journal of Financial Economics 29, 113–135.
Bayless, M., Chaplinsky, S., 1996. Is there a window of opportunity for seasoned equity offerings?
Journal of Finance 51, 253–278.
Beatty, R., Welch, I., 1996. Issuer expenses and legal liability in initial public offerings. Journal of Law
and Economics 39, 545–602.
Benveniste, L., W. Wilhelm, X. Yu, 2002. Evidence of information spillovers in the production of
investment banking services, forthcoming, Journal of Finance.
Benveniste, L., Spindt, P., 1989. How investment bankers determine the offer price in initial public
offerings. Journal of Financial Economics 24, 343–361.
Bhagat, S., Frost, P. A., 1986. Issuing costs to existing shareholders in competitive and negotiated
underwritten public utility equity offerings. Journal of Financial Economics 15, 233–259.
Booth, J. R., Chua, L., 1996. Ownership, dispersion, costly information and IPO underpricing. Journal of
Financial Economics 41, 291–310.
Booth, J. R., Smith, R. L., 1986. Capital raising, underwriting, and the certification hypothesis. Journal of
Financial Economics 15, 261–281.
Bradley, D. J., Jordan, B. D. 2002. Partial adjustment to public information on IPO underpricing. Journal
of Financial and Quantitative Analysis, forthcoming.
Busaba, W. Y., Benveniste, L. M., Guo, R., 2001. The option to withdraw IPOs during the premarket:
empirical analysis. Journal of Financial Economics 60, 73-102.
Carter, R. B., Dark, F. H., Singh, A. K., 1998. Underwriter reputation, initial returns, and the long-run
performance of IPO stocks. Journal of Finance 53, 285–311.
Carter, R. F., Manaster, S., 1990. Initial public offerings and underwriter reputation. Journal of Finance
45, 1045–1067.
Chaplinsky, S., Ramchand, L., 2000. The impact of global equity offerings. Journal of Finance 55, 2767–
2790.
Choe, H., Masulis, R. W., Nanda, V., 1993. On the timing of new equity issues: theory and evidence.
Journal of Empirical Finance 1, 3–31.
Clarke, J., C. Dunbar, K. Kahle, 2002. Long-run performance and insider trading in completed and
canceled seasoned equity offerings. Journal of Financial and Quantitative Analysis, forthcoming.
Cornelli, F., Goldreich, D., 2001. Bookbuilding and strategic allocation. Journal of Finance, 56, 2337–
2369.
Denis, D. J., 1991. Shelf registration and the market for seasoned equity offerings. Journal of Business 64,
189–212.
Dunbar, C. G., 1995. The use of warrants as underwriter compensation in initial public offerings. Journal
of Financial Economics 38, 59–78.
Eckbo, B. E., Masulis, R. W., 1991. Adverse selection and the rights offer paradox. Journal of Financial
Economics 32, 263–293.
Edelen, R.M., Kadlec, G.B, 2002. Comparable-firm returns, issuer surplus, and the pricing and
withdrawal of IPOs, manuscript, Wharton.
Fama, E. F., MacBeth, D., 1973. Risk, return, and equilibrium: empirical tests. Journal of Political
Economy 81, 607–636.
Gande, A., Puri, M., Saunders, A., 1999. Bank entry, competition and the market for corporate securities
underwriting. Journal of Financial Economics 54, 165–195.
Gerard, B., Nanda, V., 1993. Trading and manipulation around seasoned equity offerings. Journal of
Finance 48, 213–246.
Grinblatt, M., Hwang, C. Y., 1989. Signaling and the pricing of new issues. Journal of Finance 44, 393–
420.
Habib, M. A., Ljungqvist, A. P., 2001. Underpricing and entrepreneurial wealth losses in IPOs: theory
and evidence. Review of Financial Studies, 14, 433–458.
Hanley, K. W., 1993. Underpricing of initial public offerings and the partial adjustment phenomenon.
Journal of Financial Economics 34, 177–197.
Hanley, K. W., and Wilhelm, W. J., 1995. Evidence on the strategic allocation of initial public offerings.
Journal of Financial Economics 37, 239–257.
Hansen, R. S., 2001. Do investment banks compete in IPOs?: the advent of the “7% plus contract.”
Journal of Financial Economics 59, 313–346.
Hansen, R. S., Crutchley, C., 1990. Corporate financings and earnings: an empirical analysis. Journal of
Business 63, 347–371.
Heckman, J., 1979. Sample selection bias as a specification error. Econometrica 47, 153-161.
Jegadeesh, N., Weinstein, M., Welch, I., 1993. An empirical investigation of IPO returns and subsequent
equity offerings. Journal of Financial Economics 34, 153–175.
Korajczyk, R. A., Lucas, D. L., McDonald, R. L., 1991. The effects of information releases on the pricing
and timing of equity issues. Review of Financial Studies 4, 685–708.
Lease, R. C., Masulis, R. W., Page, J. R., 1991. An investigation of market microstructure impacts on
event study returns. Journal of Finance 46, 1523–1536.
Ljungqvist, A. P., Wilhelm, W.J., 2001. IPO pricing in the dot-com bubble: complacency or incentives.
Unpublished manuscript, New York University, Stern School of Business.
Loderer, C., Sheehan, D., Kadlec, G., 1991. The pricing of seasoned public offerings. Journal of Financial
Economics 29, 34–58.
Loughran, T., Ritter, J. R., 2002. Why don’t issuers get upset about leaving money on the table in IPOs?
Review of Financial Studies, forthcoming.
Loughran, T., Ritter, J. R., 2001. Why has IPO underpricing increased over time? Unpublished
manuscript, University of Florida.
Lowry, M., G. W. Schwert, 2001. Biases in the IPO pricing process. Manuscript, Penn State University.
Lowry, M., G. W. Schwert, 2002. IPO market cycles: bubbles or sequential learning? Journal of Finance,
forthcoming.
Maddala, G.S., 1983. Limited-dependent and qualitative variables in econometrics. Cambridge University
Press.
Mandelker, G., A. Raviv, 1977. Investment Banking: An economic analysis of optimal underwriting
contracts. Journal of Finance 32, 683-694.
Masulis, R. W., Korwar, A. N., 1986. Seasoned equity offerings: an empirical investigation. Journal of
Financial Economics 15, 91–118.
Megginson, W. L., Weiss, K. A., 1991. Venture capitalist certification in initial public offerings. Journal
of Finance 46, 879–903.
Mikkelson, W. H., Partch, M., 1985. Valuation effects of security offerings and the issuance process.
Journal of Financial Economics 14, 165–199.
Mikkelson, W. H., Partch, M., 1988. Withdrawn security offerings. Journal of Financial and Quantitative
Analysis 23, 119–133.
Mola, S., Loughran, T., 2001. Discounting and clustering in offer prices of SEOs, 1991–1999.
Unpublished manuscript, Bocconi University of Milan, Italy.
Myers, S. C., Majluf, N. S., 1984. Corporate financing and investment decisions when firms have
information that investors do not have. Journal of Financial Economics 13, 187–221.
Rock, K., 1985. Why new issues are underpriced. Journal of Financial Economics 15, 187–212.
Safieddine, A., Wilhelm, W. J., Jr., 1996. An empirical investigation of short-selling activity prior to
seasoned equity offerings. Journal of Finance 46, 729–749.
Sherman, A. E., Titman, S., 2002. Bookbuilding the IPO order book. Journal of Financial Economics,
forthcoming.
Smith, C. W., 1977. Alternative methods for raising capital: rights versus underwritten offerings. Journal
of Financial Economics 5, 273–307.
Smith, C. W., Jr., 1986. Investment banking and the capital acquisition process. Journal of Financial
Economics 15, 3–29.
Wilhelm, William J., 1999. Secondary market stabilization of IPOs. Journal of Applied Corporate Finance
12, 78–85.
Yeoman, J. C., 2001. The optimal spread and offering price for underwritten securities. Journal of
Financial Economics 62, 169–198.
Table 1. Underpricing, discounting, and offer-day returns for sample firms by exchange listing and as reported in prior studies. The sample is firm-underwritten, nonshelf offers by listed industrial firms from 1990 through 1997, excluding units and offers with warrants, and issues under $10 million. The rate of underpricing is the logarithm of the ratio of the offer-day closing price to the offer price. The rate of discounting is the logarithm of the ratio of the pre-offer-day closing price to the offer price. The offer-day rate of return is the logarithm of the ratio of the closing offer-day price to the prior close price. Panel A reports sample firm pricing characteristics. Closing prices are from the Center for Research in Security Prices and the offer price is from the Securities Data Company. Panel B reports comparable measures from noted studies. In parentheses are p-values, using the Student’s t-test for the means and the Wilcoxon rank test for the medians. Underpricing Discounting Offer-day return
Mean Median Mean Median Mean Median
Panel A. Sample
NYSE/Amex offers (N = 590) 1.78 0.81 1.47 0.53 0.23 0.00 (0.00) (0.00) (0.00) (0.00) (0.11) (0.12)
Nasdaq offers (N = 1,113) 3.01 1.72 3.01 2.13 -0.17 -0.29 (0.00) (0.00) (0.00) (0.00) (0.27) (0.23)
Combined 2.58 1.28 2.47 1.52 -0.03 -0.00 (0.00) (0.00) (0.00) (0.00) (0.77) (0.76)
Panel B. Prior studies
NYSE/Amex offers -0.11a 0.00b
0.12c 0.28c -0.15d
0.45e 0.44a -0.69f
0.82g 0.54g -0.70h
Nasdaq offers 1.94c 1.64c
NYSE/Amex/Nasdaq offers 0.27i 0.45e -0.29j
0.52k 0.55k
Average 0.56 0.71 -0.33
a Eckbo and Masulis (1991), 401 offers by industrial firms, 1963-1981. b Barclay and Litzenberger (1988), 218 offers by industrial firms, 1981-1983. This is an average of the prior day’s closing price to the offer-day open price and the open to the offer-day closing price. These are excess returns. c Loderer, Sheehan, and Kadlec (1991), 680 Nasdaq and 926 NYSE/Amex offers by industrial and utility firms, 1980-1984. d Lease, Masulis, and Page (1991), 210 offers by industrial and utility firms, 1981-1983. These are excess returns. e Safieddine and Wilhelm (1996), 118 offers by industrial firms, 1988-1991. f Mikkelson and Partch (1985), 311 offers by industrial firms, 1974-1983. g Smith (1977), 328 offers by industrial and utility firms, 1971-1975. h Mikkelson and Partch (1986), 56 offers by industrial firms, 1972-1982. These are excess returns. i Yeoman (2001), 1,143 offers by industrial, financial, and utility firms, 1988-1993. j Korajczyk, Lucas, and McDonald (1991), 601 over-the-counter and 646 NYSE/Amex offers by industrial firms, 1978-1983. k Safieddine and Wilhelm (1996), 356 offers by industrial firms, 1980-1987.
Table 2. Completed and withdrawn offers and the likelihood of withdrawal. The sample is firm-underwritten completed and withdrawn, nonshelf offers by listed industrial firms from 1990 through 1997, excluding units, offers with warrants, and issues under $10 million. AR(60 through 2 days before filing) is the issuer’s cumulative abnormal return from 60 days through two days before the file date. RM(file to decision day) and AR(file to decision day) are the market and issuer abnormal return during the registration period for completed offers and from the file date to the withdrawal date for withdrawn offers. Positive RM(file to decision day) and positive AR(file to decision day) are the same two variables when they are respectively positive, and zero otherwise. Amount is the proceeds. Relative amount is the ratio of shares issued to shares outstanding before the offer day. Reputation is the lead bank’s average reputation based on tombstone advertisements. Nasdaq is an indicator variable equal to one for issuers listed on the Nasdaq Exchange. Returns are constructed using prices and indices from the Center for Research in Security Prices (CRSP). In parentheses are p-values, using chi-square statistics.
Completed offers
Withdrawn offers
Mean Median Mean Median Panel A. Characteristics
AR(60 through 2 days before filing) 24.89 21.75 11.28 9.82 AR(file to decision day) -0.39 -0.92 -17.64 -16.93 RM(file to decision day) 1.26 1.08 3.02 1.76 Amount ($ millions) 85.85 51.89 62.60 44.11 Relative amount (%) 21.71 19.27 26.02 23.25 Reputation 7.82 8.75 7.42 8.75 Nasdaq (%) 65.32 1.00 70.20 1.00 N 1,707 198
Independent variable (1) (2) (3) (4)
Panel B. Probit regressions
Intercepta -0.80 -0.81 -1.09 -1.25 (0.00) (0.00) (0.00) (0.00) AR(60 through 2 days before filing) -0.71 -0.87 -0.89 (0.00) (0.00) (0.00) AR(file to decision day) -0.98 -1.18 (0.00) (0.00) Positive AR(file to decision day) 1.04 (0.00) RM(file to decision day) -0.21 -1.66 (0.58) (0.02) Positive RM(file to decision day) 1.69 (0.00) Amount -0.17 -0.16 -0.14 -0.16 (0.00) (0.01) (0.02) (0.01) Relative amount 1.13 1.26 1.55 1.52 (0.00) (0.00) (0.00) (0.00) Reputation -0.01 -0.00 0.01 0.01 (0.72) (0.98) (0.72) (0.69) Nasdaq -0.02 0.05 -0.04 -0.08 (0.82) (0.62) (0.72) (0.44) Log Likelihood 620 597 530 522 N 1,905 1,905 1,905 1,905
Table 3. Regressions of discounting on selected factors. The sample is firm-underwritten, nonshelf offers by listed industrial firms from 1990 through 1997, excluding units and offers with warrants, and issues under $10 million. Tobit estimates are reported in all models except in Column 4, which has ordinary least squares (OLS) estimates, and in Column 8, which reports Fama-MacBeth means of year-by-year regression estimates of the model. The dependent variable is the natural logarithm of the ratio of the closing price the day before the offer to the offer price. In Column 3, the dependent variable is winsorized at the 5th and 95th percentiles. Amount is the gross proceeds. Relative amount is the ratio of shares issued to shares outstanding before the offer day. 1 ÷ stock price is one over the closing stock price five days before the offer day. Volatility is the issuer’s abnormal return standard deviation for the 200 days ending one month before the offer. Nasdaq is a zero-one variable equal to one if the issuer is listed on Nasdaq. Reputation is the lead bank’s average reputation based on tombstone advertisements. Price adaptation is the ratio of the offer price to the closing price the day after the filing date. Positive price adaptation is Price adaptation when it is positive and zero otherwise. RM(registration) is the cumulative return on the Center for Research in Security Prices (CRSP) equally weighted market index during the registration period, which begins the day after the file date and ends the day before the offer date. Positive RM(registration) is RM(registration) when it is positive and zero otherwise. AR(registration) is the issuing firm’s cumulative abnormal return during the registration period, it is raw return less the CRSP equally weighted market index. Positive AR(registration) is AR(registration) when it is positive and zero otherwise. The inverse Mills’ ratio is estimated from the Probit model of offer withdrawal in Column 3 of Panel B in Table 2. An offer is primary unless it is 10% or more secondary. In parentheses are p-values, using chi-square and Student’s t-statistics.
1990s offers Independent Variable
Fama-MacBeth
Primary offers
Secondary offers
1980s offers
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) Intercept 1.00 1.01 0.01 1.02 1.07 1.01 1.02 1.04 1.02 1.02 0.99 (0.00) (0.00) (0.11) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) Amount -0.33 -0.38 -0.36 -0.58 -0.47 -0.39 -0.43 -0.47 -0.43 -0.49 -0.16 (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.02) (0.00) (0.00) (0.00) Relative amount 1.94 2.13 2.04 2.87 3.41 2.08 2.20 2.37 2.59 1.80 1.41 (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.02) (0.00) (0.08) (0.00) 1 ÷ stock price 17.10 17.28 16.79 17.15 13.18 16.80 15.89 15.65 10.61 17.28 5.50 (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.03) (0.00) (0.01) Volatility 32.49 32.73 31.03 42.40 21.55 32.46 23.75 35.41 16.99 27.35 22.76 (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.06) (0.02) (0.00) Nasdaq 0.46 0.43 0.44 0.28 0.37 0.45 0.45 0.11 0.51 0.43 0.47 (0.00) (0.00) (0.00) (0.21) (0.01) (0.00) (0.00) (0.01) (0.01) (0.09) (0.00) Reputation -0.07 -0.07 -0.06 -0.07 -0.07 -0.07 -0.06 0.06 -0.04 -0.10 -0.01 (0.01) (0.01) (0.01) (0.04) (0.01) (0.01) (0.02) (0.55) (0.30) (0.02) (0.49) Inverse Mills’ ratio -0.02 -0.02 -0.03 -0.16 -0.03 -0.05 -0.10 -0.07 -0.03 (0.12) (0.08) (0.21) (0.00) (0.07) (0.01) (0.01) (0.02) (0.39) (0.00)
continued
Table 3. continued
1990s offers Independent variable Fama-
MacBeth Tobit-primary
Tobit-secondary
1980s offers
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11)
Price adaptation -9.15 (0.00)
Positive price adaptation 10.42 (0.00)
RM(registration) -5.30 -6.60 -1.01 -7.17 -5.40 (0.29) (0.18) (0.90) (0.25) (0.46)
Positive RM(registration) 12.27 11.93 3.90 13.12 12.55 (0.09) (0.13) (0.75) (0.19) (0.23)
AR(registration) -3.31 -4.76 -2.95 -2.84 (0.00) (0.04) (0.03) (4.07)
Positive AR(registration) 6.66 6.89 6.94 5.18 (0.00) (0.01) (0.00) (0.01)
Log Likelihood 3860 3803 4147 3850 3800 3875 1772 2016 2221
Adjusted R2 0.17
N 1,703 1,703 1,703 1,703 1,703 1,703 1,703 8 818 885 764
Table 4. Discounting of early 1980s seasoned offers compared with their predicted 1990s discounting. The samples are firm-underwritten, nonshelf offers by listed industrial firms from 1980 through 1984, and from 1990 through 1997, excluding units and offers with warrants, and issues under $10 million. Amount is the gross proceeds in millions. Relative amount is the ratio of shares issued to shares outstanding before the offer day, in percent. 1 ÷ stock price is one over the closing stock price five days before the offer day. Volatility is the issuer’s abnormal return standard deviation for the 200 days ending one month before the offer, in percent. The percentage Nasdaq is the mean of a zero-one variable equal to one if the issuer is listed on Nasdaq. Reputation is the lead bank’s average reputation based on tombstone advertisements. Panel A reports mean values for the two samples and compares their means. Panel B compares the 1980s offers’ mean discounting with their predicted discounting for the 1990s using the Column 1 model from Table 3. Discounting is the logarithm of the ratio of the pre-offer-day closing price to the offer price. In parentheses are p-values using Student’s t-statistics. Variable
Amount
Relative amount
1 ÷ stock price
Volatility
Nasdaq
Reputation
Panel A. Variable means
1980s offers (n = 764) 48.72 19.62 5.19 2.79 51.96 6.93 1990s offers (n = 1,703) 85.85 21.71 4.83 3.28 65.47 7.82 Difference 37.13 2.09 -0.36 0.49 13.51 0.85 (0.00) (0.00) (0.00) (0.00) (0.00) (0.00)
By 1980s discounting group
Exchange All Under 1% 1% to 2% 2% to 3% From 3%
Panel B. Difference between 1980s issuers’ mean discounting and their predicted 1990s discounting
All offers 1980s mean 1.01 0.03 1.45 2.40 5.93 (0.00) (0.00) (0.00) (0.00) (0.00)
1990s predicted 1.94 1.56 2.38 2.62 3.27 (0.00) (0.00) (0.00) (0.00) (0.00)
Difference 0.95 1.55 0.94 0.22 -2.65 (0.00) (0.00) (0.00) (0.00) (0.00) n 764 500 136 56 72 NYSE/Amex 1980s mean 0.58 -0.06 1.39 2.34 6.31 (0.00) (0.00) (0.00) (0.00) (0.00)
1990s predicted 1.35 1.18 1.74 1.74 2.80 (0.00) (0.00) (0.00) (0.00) (0.00)
Difference 0.78 1.25 0.36 -0.58 -3.51 (0.00) (0.00) (0.00) (0.00) (0.00)
n 364 293 34 13 24 Nasdaq 1980s mean 1.40 0.15 1.47 2.42 5.74 (0.00) (0.00) (0.00) (0.00) (0.00)
1990s predicted 2.49 2.12 2.60 2.87 3.51 (0.00) (0.00) (0.00) (0.00) (0.00)
Difference 1.09 1.98 1.13 0.45 -2.23 (0.00) (0.00) (0.00) (0.00) (0.00)
N 400 207 102 43 48
Table 5. Abnormal announcement period returns and expected discounting. The sample is firm-underwritten, nonshelf offers by listed industrial firms from 1990 through 1997, excluding units and offers with warrants, and issues under $10 million. Panel A reports announcement period mean and median abnormal returns, the issuer’s raw return less the Center for Research in Security Prices (CRSP) equally weighted market index. RM(announcement) is the announcement period return on the CRSP index. Firm size is the natural logarithm of the market value of outstanding equity the day before the offer day. Relative amount is the ratio of shares issued to shares outstanding before the offer day. E[discounting] is predicted discounting by model (1) from Table 3, where the inverse of stock price five days before the announcement is used. Discrete E[discounting] is a discrete measure of discounting from 1 to 4, corresponding to the four Discrete E[discounting] groups in Panel A. An offer is primary unless it is 10% or more secondary. In parentheses are p-values, using Student’s t-statistics and Wilcoxon rank statistics. Discrete E[discounting] group
All offers
Under 1%
1% to 2%
2% to 3%
From 3%
Panel A. Announcement period abnormal returns
Mean -2.23 -1.11 -2.19 -2.49 -2.83 (0.00) (0.00) (0.00) (0.00) (0.00)
Median -2.24 -1.45 -2.54 -2.60 -2.63 (0.00) (0.00) (0.00) (0.00) (0.00)
N 1703 321 481 518 383
All offers
Primary offers
Secondary offers Independent
variable (1) (2) (3) (4) (5)
Panel B. Regressions of the announcement period return
Intercept 0.26 -0.24 -0.07 -0.05 -0.09 (0.04) (0.24) (0.10) (0.45) (0.11)
RM(announcement) 1.22 1.21 1.23 1.27 1.19 (0.00) (0.00) (0.00) (0.00) (0.00)
Firm size 0.62 0.27 0.20 0.35 (0.00) (0.15) (0.47) (0.18)
Relative amount -0.47 1.36 1.42 0.43 2.74 (0.71) (0.32) (0.30) (0.84) (0.13)
E[discounting] -0.27 0.09 (0.03) (0.60)
Discrete E[discounting] -0.42 -0.57 -0.29 (0.02) (0.02) (0.09)
Adjusted R2 0.03 0.03 0.04 0.04 0.03
N 1,703 1,703 1,703 818 885
Table 6. Daily stock abnormal return variance ratios for 11 days centered on the offer day. The sample is firm-underwritten, nonshelf offers by listed industrial firms from 1990 through 1997, excluding units and offers with warrants, and issues under $10 million. The variance ratio is the squared difference of the daily common stock rate of return, except on the offer day when the return is adjusted for the dilution from the discounting surprise, less the corresponding market return, to the issuer’s benchmark daily abnormal return variance. For all issues, the benchmark is the variance of all returns for the 200 days ending one month before the offer. For issuers with positive returns on the relative day, the benchmark is the variance of all positive returns for the same period. In parentheses are p-values using Student’s t-statistics, followed by number of observations.
All issuers
Issuers having positive returns on the relative day
All NYSE/Amex Nasdaq All NYSE/Amex Nasdaq
Day relative to the offer day (1) (2) (3) (4) (5) (6)
-5 0.97 1.11 0.95 1.06 1.11 1.03 (0.82) (0.29) (0.51) (0.46) (0.44) (0.79) 1703 590 1113 707 256 451
-4 0.99 0.98 0.99 1.14 1.14 1.13 (0.95) (0.82) (0.92) (0.14) (0.43) (0.42) 1703 590 1113 656 208 448
-3 1.08 1.03 1.10 1.07 1.10 1.05 (0.21) (0.71) (0.22) (0.43) (0.54) (0.61) 1703 590 1113 656 226 430
-2 1.15 0.98 1.23 1.02 0.87 1.09 (0.01) (0.80) (0.00) (0.84) (0.16) (0.20) 1703 590 1113 599 204 395
-1 1.26 1.34 1.22 1.04 1.08 1.02 (0.00) (0.02) (0.00) (0.57) (0.58) (0.75) 1703 590 1113 602 196 406
Offer 1.70 1.95 1.56 1.86 2.07 1.73 (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) 1703 590 1113 740 271 469
1 0.84 0.92 0.80 1.02 1.19 1.01 (0.00) (0.39) (0.00) (0.99) (0.21) (0.92) 1703 590 1113 803 284 519
2 0.85 1.02 0.76 0.96 1.05 0.97 (0.00) (0.84) (0.00) (0.55) (0.66) (0.75) 1703 590 1113 777 260 517
3 0.85 1.12 0.71 0.92 1.15 0.85 (0.01) (0.36) (0.00) (0.52) (0.34) (0.06) 1703 590 1113 778 278 500
4 0.88 0.95 0.84 1.01 0.97 1.10 (0.01) (0.63) (0.01) (0.38) (0.84) (0.36) 1703 590 1113 761 260 501
5 0.85 0.95 0.80 0.92 0.97 0.92 (0.00) (0.51) (0.00) (0.98) (0.82) (0.35) 1703 590 1113 743 266 477
Table 7. Regressions of daily stock returns on and around the offer day on surprise discounting. The sample is firm-underwritten, nonshelf offers by listed industrial firms from 1990 through 1997, excluding units and offers with warrants, and issues under $10 million. Reported are ordinary least squares tests. The dependent variable is the logarithm of one plus the daily stock rate of return, adjusted for the dilution from the discounting surprise, using the Center for Research in Security Prices (CRSP) closing stock prices, except in Column 1 where it is the logarithm of one plus the daily stock rate of return. Columns 6 through 10 report results from regressions on each of five days centered on the offer day. Surprise is the residual from the Column 7 model for discounting in Table 3. Fama-MacBeth surprise is the discounting less the discounting predicted using the Column 8 Fama-MacBeth coefficient estimates of Table 3. The symbol x indicates the product of the two associated variables to form a multiplicative interaction variable. AR(ann.) is the abnormal return on the issuer’s common stock at the announcement of the offering. Surprise I, II, IV, and V are four, zero-one variables, respectively equal to one if the surprise is in the 1st, 2nd, 4th, and 5th quintile of the surprise frequency distribution. RM is the equally weighted offer-day return on the CRSP index. Firm size is the natural logarithm of the market value of outstanding equity the day before the offer-day. Abnormal turnover is offer day trading volume relative to the daily average trading volume over the year after 150 days after the offer. The inverse Mills’ ratio is estimated from the Probit model of offer withdrawal in Column 3 of Panel B in Table 2. In parentheses are p-values using Student’s t-statistics, followed by number of observations. Five days centered on the offer day Independent Offer day Day -2 Day -1 Offer day Day +1 Day +2
variable (1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
Intercept -2.57 8.47 -3.61 -3.09 -2.57 -4.73 5.13 -3.47 -1.52 3.01 (0.23) (0.00) (0.01) (0.16) (0.24) (0.02) (0.01) (0.13) (0.35) (0.07)
Surprise -0.62 -0.61 -0.62 (0.00) (0.00) (0.00)
Discounting -0.57 (0.00)
Fama-MacBeth surprise -0.59 (0.00)
Surprise x firm size -0.00 (0.29)
Surprise x AR(ann.) 0.06 (0.91)
continued
Table 7. continued
Five days centered on the offer day Independent Offer day Day -2 Day -1 Offer day Day +1 Day +2
variable (1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
Surprise I -0.10 -1.09 1.21 0.05 -0.08 (0.72) (0.00) (0.00) (0.81) (0.71)
Surprise II 0.11 -0.29 0.55 0.20 0.o9 (0.70) (0.29) (0.06) (0.34) (0.65)
Surprise IV -0.11 0.33 -1.21 0.12 0.40 (0.70) (0.22) (0.00) (0.56) (0.07)
Surprise V -0.03 0.03 -3.13 0.12 0.37 (0.92) (0.92) (0.00) (0.58) (0.10)
RM 1.52 1.55 1.51 1.51 1.52 1.17 1.57 1.56 -0.12 -0.24 (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.37) (0.07)
Firm size 0.19 -0.28 0.23 0.22 -0.19 0.29 -0.04 0.26 0.06 -0.14
(0.03) (0.00) (0.01) (0.02) (0.03) (0.00) (0.64) (0.01) (0.39) (0.04)
Abnormal turnover -1.80 -2.34 -1.65 -1.77 -1.80 -0.08 -0.27 -1.39 0.02 -0.01 (0.21) (0.10) (0.25) (0.22) (0.21) (0.51) (0.00) (0.34) (0.82) (0.21)
Inverse Mills’ ratio -0.03 -0.05 -0.08 -0.03 -0.03 -0.03 -0.15 -0.03 0.02 -0.00 (0.22) (0.06) (0.00) (0.23) (0.22) (0.22) (0.00) (0.20) (0.21) (0.79)
Adjusted R2 0.18 0.18 0.18 0.18 0.18 0.04 0.08 0.16 -0.00 0.01
N 1,702 1,702 1,702 1,702 1,702 1,702 1,702 1,702 1,702 1,702
Table 8. Regressions of daily stock returns on surprise discounting. The sample is firm-underwritten, nonshelf offers by listed industrial firms from 1990 through 1997, excluding units and offers with warrants, and issues under $10 million. In columns 4-6, offers are included only if their underpricing is above -2%, 0%, and 2%, respectively. Underpricing is the logarithm of the ratio of the offer-day closing price to the offer price. Reported are ordinary least squares regression coefficients. The dependent variable is the logarithm of one plus the offer day stock rate of return, adjusted for the dilution from the discounting surprise, using the Center for Research in Security Prices (CRSP) closing stock prices. Surprise is the residual from the Column 7 model for discounting in Table 3. Good surprise equals surprise if Surprise is negative (i.e., good news) and zero otherwise. RM is the equally weighted offer-day return on the CRSP index. Firm size is the natural logarithm of the market value of outstanding equity the day before the offer day. Abnormal turnover is offer day trading volume relative to the daily average trading volume over the year after 150 days after the offer. The inverse Mills’ ratio is estimated from the Probit model of offer withdrawal in Column 3 of Panel B in Table 2. An offer is primary unless it is 10% or more secondary. In parentheses are p-values using Student’s t-statistics, followed by number of observations. All Primary Secondary All offers with underpricing above
Independent offers offers offers - 2% 0% 2%
variable (1) (2) (3) (4) (5) (6)
Intercept -4.03 -2.20 -6.29 -1.62 2.04 2.21 (0.07) (0.49) (0.06) (0.45) (0.42) (0.94)
Surprise -0.55 -0.68 -0.42 -0.53 -0.58 -0.57 (0.00) (0.00) (0.00) (0.00) (0.00) (0.00)
Good surprise -0.29 -0.32 -0.34 -0.36 -0.25 -0.23 (0.04) (0.13) (0.06) (0.01) (0.09) (0.16)
RM 1.54 1.70 1.35 1.22 0.88 0.60
(0.00) (0.00) (0.00) (0.00) (0.00) (0.03)
Firm size 0.26 0.12 0.42 0.14 -0.00 0.17 (0.01) (0.37) (0.00) (0.10) (0.93) (0.19)
Abnormal turnover -1.68 -2.93 -1.25 -1.53 -0.78 0.03 (0.24) (0.13) (0.56) (0.27) (0.63) (0.89)
Inverse Mills’ ratio -0.03 0.01 -0.07 -0.02 0.00 -0.01 (0.20) (0.82) (0.03) (0.32) (0.99) (0.76)
Adjusted R2 0.18 0.20 0.17 0.20 0.23 0.27
N 1,702 885 818 1,592 1,067 721
Figure 1. Frequency distributions of discounting, underpricing, and offer-day return. The sample is firm-underwritten, nonshelf offers by listed industrial firms from 1990 through 1997, excluding units and offers with warrants, and issues under $10 million. The rate of underpricing is the logarithm of the ratio of the offer-day closing price to the offer price. The rate of discounting is the logarithm of the ratio of the pre-offer-day closing price to the offer price. The offer-day rate of return is the logarithm of the ratio of the offer-day closing price to the pre-offer-day closing price.
0%
5%
10%
15%
20%
25%
-20% -15% -10% -5% 0% 5% 10% 15% 20%
Panel A. Rate of underpricing
0%
5%
10%
15%
20%
25%
-20% -15% -10% -5% 0% 5% 10% 15% 20%
Panel B. Rate of discounting
0%
5%
10%
15%
20%
25%
-20% -15% -10% -5% 0% 5% 10% 15% 20%
Panel C. Offer-day rate of return
Figure 2. The rate of discounting versus several issuer characteristics. The sample is firm-underwritten, nonshelf offers by listed industrial firms from 1990 through 1997, excluding units and offers with warrants, and issues under $10 million. The variables are defined in Table 3. The mean and median discounts are measured within the respective quartiles (lowest I to highest IV) or in the case of exchange listing, by exchange. Discounting is the logarithm of the ratio of the pre-offer-day closing price to the offer price.
0%
1%
2%
3%
4%
I II III IVPanel A. Amount
0%
1%
2%
3%
4%
I II III IVPanel B. Relative amount
0%
1%
2%
3%
4%
I II III IVPanel C. 1÷stock price
0%
1%
2%
3%
4%
I II III IVPanel D. Volatility
0%
1%
2%
3%
4%
Yes NoPanel E. Nasdaq
0%
1%
2%
3%
4%
I II III IVPanel F. Reputation
Figure 3. Discounting versus the offer price adaptation, market returns, and abnormal returns. The sample is firm-underwritten, nonshelf offers by listed industrial firms from 1990 through 1997, excluding units and offers with warrants, and issues under $10 million. The registration period starts the day after the Securities and Exchange Commission file date and ends the day before the offer. The offer price adaptation is the logarithm of the ratio of the offer price to the closing price on the day after the file date. The market return is the registration period rate of return on the Center for Research in Security Prices (CRSP) equally weighted index. The registration period abnormal return is the issuer’s raw return less the return on the CRSP equally weighted index over the registration period. Discounting is the logarithm of the ratio of the closing price the day before the offer to the offer price. The three variables are separately sorted from low to high into deciles I-X, which are reported along with their within decile means.
0%
1%
2%
3%
4%
5%
6%
I-29.0%
II-14.8%
III-9.8%
IV-6.5%
V-3.5%
VI-0.5%
VII2.2%
VIII5.7%
IX11.1%
X27.3%
Panel A. Offer price adaptation
0%
1%
2%
3%
4%
I-4.0%
II-1.7%
III-0.7%
IV-0.0%
V-0.7%
VI1.5%
VII22.2%
VIII3.1%
IX4.3%
X7.3%
Panel B. Market rate of return
0%
1%
2%
3%
4%
I-26.7%
II-12.7%
III-8.3%
IV-5.0%
V-2.4%
VI0.2%
VII3.0%
VIII6.6%
IX12.2%
X27.2%
Panel C. Issuers’ abnormal return
Figure 4. Plot of the offer-day rates of return versus the surprise in the rate of discounting. The sample is firm-underwritten, nonshelf offers by listed industrial firms from 1990 through 1997, excluding units and offers with warrants, and issues under $10 million. The surprise discounting is the residual from the Column 1 model for discounting reported in Table 3. The offer-day return is the logarithm of the ratio of the offer-day closing price to the closing price the day before.
-20%
20%
-20% 20%
Discount surprise
Offer-day return